A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 06:00 UTC
[Doc] Update description disable_any_whitespace (#29784)
Signed-off-by: Frederic Odermatt <frederic.odermatt@44ai.ch>
FredericOdermatt committed
Dec 1, 2025 at 17:48 UTC
5d43f7372ebe0e306a03142d3815f76ff0147240
1 file changed
+4
-2
vllm/config/structured_outputs.py
+4
-2
@@ -28,8 +28,10 @@ class StructuredOutputsConfig:
28
disable_fallback: bool = False
29
"""If `True`, vLLM will not fallback to a different backend on error."""
30
disable_any_whitespace: bool = False
31
- """If `True`, the model will not generate any whitespace during structured
32
- outputs. This is only supported for xgrammar and guidance backends."""
31
+ """If `True`, json output will always be compact without any whitespace.
32
+ If `False`, the model may generate whitespace between JSON fields,
33
+ which is still valid JSON. This is only supported for xgrammar
34
+ and guidance backends."""
35
disable_additional_properties: bool = False
36
"""If `True`, the `guidance` backend will not use `additionalProperties`
37
in the JSON schema. This is only supported for the `guidance` backend and